- Email: [email protected]
silica catalysts
213
Applied Catalysis, 70 (1991) 213-224 Elsevier Science Publishers B.V., Amsterdam
Arsine poisoning
of nickel/silica
Hydrogen chemisorption
catalysts
study by magnetic
method
C.F. Ng’ Department
of Chemistry> Hong Kong Baptist
College, 224 Waterloo
Road, Kowloon
(Hong Kong)
and Y.J. Chang Department
of Chemistry,
(+852)3397011,
fax. (+852)
CYnivers@
of Hong Kong, Pokfulam
3388014, E~mail cfng@&7856.
Road (Hong Kong),
tel.
hkbc.hk
(Received 4 July 1990, revised manuscript received 2 November 1990)
Abstract The poisoning of Ni/SiO, catalysts by arsine adsorption and the subsequent chemisorption of hydrogen were studied by the Weiss extraction method. Arsine was found to adsorb dissociatively on the nickel catalyst surface with evolution of hydrogen at a coverage of 8=0.7. Many layers of nickel metal may be attacked. The amount of surface accessible to hydrogen adsorption is considerably less than that expected on the basis of a simple ‘blocking mechanism’; this leads to the postulate of an arsenic atom being bonded to three nickel atoms while at the same time interacting with a second group of nickel sites sufficiently strongly to cause them to become inaccessible to chemisorption (but not enough to prevent them from participating in collective ferromagnetism). This finding seems to be the first piece of experimental evidence that a poison atom may exert its influence over host atoms differentially. An czAe value as high as 8, found after heat treatment at ca. ZOO”C, suggests the formation of clusters of AsNi, with n being as high as 12. Such clusters may change to As,Ni, when the amount of arsenic increases in the nickel catalyst. The electronic structure of the AsNi,, cluster is expected to be very similar to that of Ni,,, which was treated by Messmer et al. In the light of the present work, the cluster model approach might provide a unifying basis for the understanding high of the magnetic effect of carbon, sulphur and arsenic on nickel catalysts that are subject to heat treatment. Keywords: arsine poisoning, magnetic method, nickel/silica,
adsorption, cluster model.
INTRODUCTION
While sulphur poisoning has been the subject of extensive investigation in recent years, arsenic poisoning studies are relatively scarce, despite its relevance in industrial catalytic processes. In this paper we report our findings on the study of the arsine poisoning of Ni/SiO, catalysts and the subsequent chemisorption of hydrogen using the Weiss extraction method. The results are
0166-9834/91/$03.50
0 1991 Elsevier Science Publishers
B.V.
214
also discussed in the light of our earlier findings on the hydrogen sulphide poisoning of nickel catalysts. EXPERIMENTAL
The method of preparation of Ni/SiO, catalysts, the apparatus set-up and the procedure for adsorption and desorption experiments were similar to those outlined in a previous paper [ 11. The average nickel loading value and nickel particle diameter in the catalysts so obtained were lo-12% and 5.3 nm, respectively. Arsine was prepared by reducing As (OH), (aq) by potassium tetrahydroborate according to the method of Jolly and Drake [ 21. Magnetic data were obtained with an automatic Weiss extraction apparatus which has been described in detail elsewhere [ 31 s The value of bond number n was calculated by dividing ~1,the change in saturation magnetization per adsorbed molecule or atom, by the magnetic moment of nickel atoms, i.e. n= [B.M./molecule] / 0.606 B.M. RESULTS
AND DISCUSSION
Nature of chemical state of arsine The shape of the magnetic isotherm of arsine adsorption on Ni/SiOz shown in Fig. 1 suggests that the chemical nature of adsorbed arsine changes at different coverages. Obviously, three regions may be recognized. In region I, V,,, = O-8 ml, there is a marked decrease of magnetization of the catalyst. Although this part of the isotherm is somewhat curved, it is still possible to estimate an average value of m 4.0 B.M. - equivalent to a bond number of 6 by drawing the dotted line as shown. In region III, viz. V,,, = 17-36 ml, a very flat decrease in magnetization is effected by arsine adsorption. The region between the two could be considered to be a transition region. It is also interesting to point out that hydrogen evolution was observed beyond point E and that point M ( VA~H~ = 15 ml) corresponds to monolayer coverage when compared with the results of earlier hydrogen adsorption studies [ 1,5]. These results are therefore consistent with an initial dissociation adsorption of arsine with arsenic probably being bonded to three nickel sites, while the three hydrogen atoms are bonded to another three nickel sites, as shown in the following sketch: , N’l
/
As \ : NI
H \
\
\ Ni
H
H (1)
Nl
Ni
Ni
From a thermodynamic point of view, this model is well supported since the enthalpy change for the reaction ASH, + NiLNiAs + 3.2 H2 is - 278.7 kJ/mol. That desorption of arsine still takes place beyond point M indicates that ar-
215
\B c
A
A
30-
-
-_
t
I
0
I
4
1
8
1
12
1
16 20
1
24
1
28 32 36
I
I
1
40
Fig. 1. Change in magnetization as a function of adsorbed volume at room temperature for arsine. Curve A: for H,; Curve B is based on hydrogen adsorption results of Martin et al. [ 5 1,whose slope yields a”2 = 1.4 B.M./molecule. Point M represents monolayer coverage. Point E represents the volume adsorbed
at which hydrogen gas evolution
occurs.
senic has attacked more than the top layer of nickel atoms, although the bond number due to amine is apparently much lower when deeper layers are involved. This phenomenon is qualitatively very similar to that found by Ng and Martin [l] for hydrogen sulphide adsorption, where it was deduced that hydrogen sulphide was bonded to four nickel atoms of the first layer and approximately to two nickel atoms of the second layer. There are differences between arsine and hydrogen sulphide adsorption, however, in that the former has a transition region and hydrogen evolves much earlier - before monolayer coverage is reached ( tlAAsH3 = 0.7 ) , whereas in the latter case virtually no transition region is present and hydrogen evolution occurs only after two layers of nickel are covered (&,, = 2.0). Furthermore, the very existence of the transition region and the fact t.hat hydrogen evolution does not occur until half-way through this region imply that some less dissociated species, such as
216
H I
H\\ or
AS\
/, \
H
As'
\
Ni”
‘Ni (II)
Nl (III)
may be present when more than half a monolayer is covered by arsine. Structures II and III have been suggested earlier by Al-Daher and Saleh [ 41 in their studies of arsine adsorption on metallic nickel films at - 80’ C. However, it should be pointed out that their deduction was based on the assumption that only the very top layer of the nickel surface would be attacked. Finally, arsine adsorption differs from that of hydrogen sulphide in one more aspect: adsorption of arsine was found to occur even after the magnetization value of the nickel catalyst sample dropped to zero (not shown in Fig. 1) suggesting that much deeper layers were involved, leading to the formation of bulk Ni,As, which does not possess a superparamagnetic property. Nature of chemical states of arsenic in (on) nickel Curves B and C in Fig. 2 shows that the volume of hydrogen released at a given temperature is higher with greater arsine coverage. If we take curve A into consideration as well, it appears that hydrogen evolves more effectively on an arsine covered surface than on a hydrogen covered surface. Futhermore, all hydrogen could be completely removed from the surface at 673 K. These observations are quite different from that of the hydrogen sulphide case, where the presence of sulphur suppresses the release of hydrogen and a considerable amount of hydrogen is retained in (on) the poisoned catalyst, even at 773 K. Thus it appears that, in contrast to sulphur, arsenic has a promoting effect on hydrogen evolution. Alternatively speaking, arsenic weakens the Ni-H bond and this therefore implies that the effect of arsenic is not confined to the sites to which it is directly bonded. This deduction is further supported by the results of hydrogen adsorption on a poisoned surface (Fig. 5). That the an2 value remains almost the same as that for a bare nickel surface up to 8 ml [NTP]/g Ni of arsine coverage, as shown in Fig. 3, and the straight line yielding an2 = 1.3 for hydrogen readsorption, shown in Fig. 4, indicate that the uncovered part of a partially arsine poisoned surface accessible to hydrogen acts as a ‘free’ surface toward hydrogen adsorption. It would therefore be expected that, as in the case of hydrogen sulphide, the corresponding VH2vs* V.4, HB P lot should be a straight line up to the transition region (dotted line in Fig. 5 ) because only a blocking mechanism is involved. The experimentally observed curve falls short of this expectation. The maximum uptake is always lower than the value expected from a simple blocking mechanism and
217
Fig. 2. Amount of desorbed hydrogen
( Vn, per V nP adsorbed as Hz or as ASH,) from Ni/SiO,
with
various arsine coverages as a function of temperature [A: Ni covered with H,; B: Ni covered with 2.9 ml(NTP)AsH,/g Ni; C: Ni covered with 7.1 ml(NTP)AsH,/g Nil.
indeed almost reaches zero at 7-8 ml/g Ni amine coveragea and is completely zero at about 10 ml/g Ni ASH, coverage, a value that corresponds to the commencement of hydrogen evolution on an arsine covered surface. These experimental results can only be explained by postulating that on a partially arsine covered surface, some parts of the bare surface (though bare) are not accessible to further hydrogen adsorption and those parts which are accessible behave as ‘free’ surface as far as hydrogen adsorption is concerned. In fact, one can calculate the V,, vs. VA, H3 relationship on the basis of a model which envisages that each adsorbed arsenic atom, apart form directly binding three nickel sites, will interact fairly strongly with five more sites so that altogether eight sites will not be accessible to adsorption of a gas that requires free surface sites (although the five additional nickel atoms will still participate in collective ferromagnetism) ~ Such a theoretical relationship is shown as Curve C in Fig. 5 and should be compared with dashed line B which is expected on the basis of a simple blocking mechanism based on species I OThe value of i-8 ml NTP/g as carbon monoxide,
Ni is also the same value of coverage beyond which other gases such
ethylene. acetylene
or propene,
studied in this work, cannot be adsorbed.
218
t L
0
1
I
2
3
4
5
Fig. 3. cxH, vs. the volume of preadsorbed
6
7
8
ijA,E3(ml
NTP/gNi)
arsine.
throughout region I of arsine adsorption. (It should be noted that the shape of the dashed line B beyond the region I of arsine adsorption is ill defined.) The closeness between curves C and A appears to lend support to this model. Such a model is indeed in line with the thinking that a poison atom may have a detrimental influence beyond its nearest neighbours, which has been adopted by various authors in discussing gas adsorbate that requires multiple sites (see e.g. Andersen et al. [ 671) Martinet al. [ 8,9] ). To the knowledge of the present authors, our results are the first piece of direct experimental evidence which shows that, firstly a poison atom can exert an influence on two sets of atoms differentially and, secondly, the inaccessibility of the second set of atoms is definitely not due to any ‘steric effect’ since the adsorbate in question is hydrogen which does not require multiple sites. On the basis of this finding, one may expect that the poisoning effect of arsenic should be more severe than that of sulphur, at least for reactions involving hydrogen. This expectation is quite different from the results reported by Nielsen and Villadsen [lo]. The chemical states of arsenic in (on ) the nickel surface heat treatment is also of interest. Fig. 6 shows t,hat the cy value of a sample (after subtracting (xu due to remaining surface hydrogen, if any) is considerably higher after hydrogen evolution at elevated temperature (measurements were made after
219
Fig. 4. Change in magnetization
of a nickel catalyst treated with 6.6 ml of ax-sine at room temper-
ature and outgassed at 450’ C as a function an (Y value of 1.4 B.M./molecule.)
of
V,, readsorbed at room temperature.
(Slope yields
cooling down). An cyA,value as high as 8.0 (i.e. bond number = - 12-13) was observed for a sample of low arsine coverage and o!yAs was found to decrease as the amount of As increased. Such a high value is indeed astonishing because the arsenide with the highest Ni/As ratio reported is Ni5Asz. Furthermore, if one adopts the approach of Martin et al. [ 111 in interpreting the reduction of magnetic moment of a nickel catalyst by interstitial carbon, then the maximum number of electrons that can be donated to the d-band of metallic nickel by an interstitial arsine atom is five, i.e. the maximum ayAsvalue should be five, rather than the eight found experimentally. A more plausible explanation for the large observed bond number may be due to the formation of AsNils clusters, which results from the thermally induced migration of a surface arsenic atom during heat treatment to displace one of the bulk nickel atoms (most likely in the second or third layer below the surface). Imagine that this arsenic atom decreases the density of states at Fermi level energy (n(E,) ) of ca. 12 neighbouring nickel atoms in such a way that the well known Stoner criterion of ferromagnetism, UX n (I&.)> T (U,the exchange interaction energy; T the temperature in question) is not satisfied, then macroscopically, a decrease in magnetization will be observed which corresponds to 12 nickel atoms not par-
Fig. 5. Volume of adsorbed hydrogen at room temperature vs. the volume of preadsorbed amine. A: experimental data; B: expected curve if simple blocking mechanism operates with the same adsorbed species (Species I) throughout the first region as arsine adsorption. C: calculated on the basis of presently proposed model of an arsenic atom binding to three nickel sites and interacting fairly strongly with a second group of five nickel sites.
ticipating in collective ferromagnetism. Another approach to explaining the magnetic property arising from the AsNi,, cluster is based on the cluster model. Such a cluster is very similar to the cube-octahedral Nils cluster whose electronic structure has been calculated and fully discussed by Messmer et al. [ 121. According to these authors, amongst the spin polarized outer energy levels, 68 and 62 electrons of the nickel atoms (nickel has 10 outermost electrons, 3ds4s2) will occupy the spin-up and spin-down levels, respectively, leaving a total of six unpaired electrons and thereby leading to a net spin polarization and paramagnetism of the cluster. Now, in view of the close proximity of nickel and arsine in electronegativity and size (;cNi= 1.8, J&= 2.0; rr+Ji =0.125 nm, r&,=0.121 nm), it is conceivable that the substitution of the central nickel atom by an arsine atom is realistic and that the resulting AsNi,, cluster will presumably have an electronic structure very similar to that of a Ni,, cluster, except for the fact that the five extra electrons brought in by the arsenic atom ( 3d1°4s2p3) will occupy the spin-down levels, so that the AsNil cluster has only one paramagnetic spin, rather than six as in the case of Nil3 cluster. To a
221
90
80 . 70. 60. 50. 40. 30. 201 0
373
473
573
673
T( K)
Fig. 6. cy as a function of the desorption temperature for arsine. [A, B and C correspond, tively, to VAsH3= 1.3,3.8 and 8.2 ml NTP/g Ni.]
respec-
first approximation, the expected effect of the formation of such an AsNi,, cluster on magnetization of the nickel catalyst should be a reduction corresponding to 12 nickel atoms being hindered from participating in collective ferromagnetism. The experimental results are not far from this predicted reduction. It is interesting to note that, apart from the high magnetic effect reported for carbon, as mentioned above, Ng and Martin [1 ] also found that, at elevated temperature, a sulphur atom could decrease the magnetization of nickel by as much as 4 B.M. (corresponding to 6-7 nickel atoms being hindered from contributing to collective ferromagnetism). In the light of the present work, it appears that the cluster model might provide a unifying explanation for these phenomena. Both carbon and sulphur differ significantly from nickel in terms of electronegativity, size, and number of electrons in the outer shells and are therefore very unlikely to form an XNi,, (X = C, S ) cluster”. Instead, a XNis cluster may more likely be formed when the nickel catalyst (with X adsorbed) is thermally treated. Theoretical calculations of energy levels for AsNi,,, CNis and SNiGwould be very useful to confirm the feasibility of the proposed AsNil cluster and the general validity of the cluster model in interpreting the ob“For a close packed structure, it is possible to retain ligancy 12 with a central sphere as much as 10% smaller than the surrounding spheres. These spheres are then arranged at the corners of a regular icosahedron. See e.g. Pauling [ 131.
222
served high magnetic effects of carbon, sulphur and arsenic on nickel catalysts subjected to heat treatments. The heat treatment results obtained, as shown in Fig. 7, provide similar information. (Here the sample was not evacuated and the hydrogen desorbed should be readsorbed onto the surface after cooling.) Perhaps two more points should be noted. First, for a surface with greater amount of arsenic, both Fig. 6 and Fig. 7 show that the aAs value is much smaller, suggesting formation of cluster-like As,Ni, (or even bulk arsenides, as mentioned earlier) at high coverage. Second, in comparing Fig. 6 with Fig. 7, the presence of hydrogen has very little effect on the value of a& ;IZ agreement with our deduction in the discussion of hydrogen adsorption presented earlier. By comparison with our previous studies of hydrogen sulphide, one can see that the effect of As on catalytic reactions in general (such as partial hydrogenation of acetylene to ethylene, of polyalkenes or cyclic polyenes to corresponding monoalkenes) will be expected to be more severe than that of sulphur. In our next communication we shall report the effects of arsenic on the adsorption of carbon monoxide, acetylene, ethylene and propene. In comparing the results obtained for carbon monoxide and acetylene with those of hydrogen sulphide arsine effects are more complex, but no doubt render the sura AS~;3(B. M. /molecule)
373
473
c 573
Fig. 7.LYas a function of the holding temperature spectively, to VA, H3 =0.85, 1.8, 7.1 ml NTP/g Ni.
T( K)
673 for arsine. Curves A, B and C correspond
re-
223
face significantly less accessible to these gases (the ratio of maximum uptake of carbon monoxide for arsenic to sulphur is 1: 3 ). As regards acetylene, a completely sulphur poisoned nickel surface was found still to be capable of adsorbing, while a surface with - 7 ml AsH,/g Ni was found to be incapable of adsorption of acetylene and was therefore more ‘poisonous’ for reactions involving either the adsorbed carbon monoxide or acetylene species as precursor (or adsorption of carbon monoxide or acetylene as a rate determining step). Insofar as hydrocarbon cracking is concerned, in a broad sense, the inhibiting power of arsenic was found to be comparable with that of sulphur. However, the effect was again found more complex in the case of arsenic and this will be also discussed in our next communication [ 141. CONCLUSION
Arsenic can attack many layers of a nickel catalyst at room temperature. Before hydrogen evolution at 0% 0.7, the chemical state of arsine in (on) the nickel catalyst is probably H H H I AS\ \ I
N;
Ni
‘Xi
Ni
Ni
Ni
An arsenic atom, apart form bonding to a first group of nickel atoms, can interact with a second group of nickel sites causing them to be inaccessible to hydrogen adsorption. Such an interaction, however, is not strong enough to prevent the second group of nickel sites from participating in collective ferromagnetism. For those parts of the nickel surface which are available for hydrogen adsorption, they behave as free surface. This finding seems to be the first piece of experimental evidence that a poison atom may exert its influence over host atoms differentially. The chemical state of arsenic undergoes changes at elevated temperature. The very high aAs value observed when the amount of arsenic in a sample is small could be accounted for by postulating the formation of an AsNil cluster arising from thermally induced migration of a surface arsenic atom to replace a bulk nickel atom. The electronic structure of such a cluster is expected to be very similar to that of NilY cluster which has been studied by Messmer et al. [ 121. When the amount of arsenic increases, the number of nickel atoms participating in bonding decreases, suggesting formation of a cluster such as As,Ni, at higher coverage. The behaviour of the poisoned surface towards adsorption suggests that the poisoning effect of arsenic should be more severe than that of sulphur, at least for reactions involving hydrogen. ACKNOWLEDGEMENT
We should like to thank Messrs. C.P. Luk, F. Lam and Miss Jennifer Lie of the Department of Chemistry, University of Hong Kong for their assistance
224
in the technical work. One of us (Y.J. Chang) is indebted to the Pei Hua Education Foundation of Hong Kong for a Fellowship. Helpful comments from Dr. G.A. Martin are also gratefully acknowledged.
REFERENCES
2 3 4 5 6 7 8 9 LO 11 12 13 14
C.F. Ng and G.A. Martin, J. Catal., 54 (1978) 384. J.E. Drake and C. Riddle, J. Inorg. Synt., 13 (1957) 14. Y.J. Chang, C.F. Ng and W.K. Kwan, Rev. Sci. Instru., 59 (1988) 342. I.M. Al-Daher and J.M. SaIeh, J. Phys. Chem., 76 (19’72) 2851 and references therein. G.A. Martin, B. Imelik and M. Pettre, J. Chim. Phys., 66 (1969) 1682. N.T. Andersen, F. Topsee, I. Alstrup and J.R. Rostrup-Nielsen, J. Catal., 104 (1987) 454. I. Alstrup and N.T. Andersen, J. Catal., 104 (1987) 466. G.A. Martin, Catal. Rev. - Sci. Eng., 30 (1988) 519. G.A. Martin and CF. Ng, Appl. Catal., 31 (1987) 235. B. Nielsen and J. Villadsen, Appl. Cat& 11 (1984) 123. G.A. Martin, M. Primet and J.A. Dalmon, J. Catal., 53 (1978) 321. R.P. Messmer, S.K. Krmdson, K.J. Johnson, J.B. Diamond and C.Y. Yang, Phys. Rev. B, 13 (1976) 1396. L. Pauling, The Nature of the Chemical Bond, 3rd Ed., Cornell Univ. Press, 1980. C.F. Ng and Y.J. Chang, to be submitted to Appl. Catal.
Applied Catalysis, 70 (1991) 213-224 Elsevier Science Publishers B.V., Amsterdam
Arsine poisoning
of nickel/silica
Hydrogen chemisorption
catalysts
study by magnetic
method
C.F. Ng’ Department
of Chemistry> Hong Kong Baptist
College, 224 Waterloo
Road, Kowloon
(Hong Kong)
and Y.J. Chang Department
of Chemistry,
(+852)3397011,
fax. (+852)
CYnivers@
of Hong Kong, Pokfulam
3388014, E~mail cfng@&7856.
Road (Hong Kong),
tel.
hkbc.hk
(Received 4 July 1990, revised manuscript received 2 November 1990)
Abstract The poisoning of Ni/SiO, catalysts by arsine adsorption and the subsequent chemisorption of hydrogen were studied by the Weiss extraction method. Arsine was found to adsorb dissociatively on the nickel catalyst surface with evolution of hydrogen at a coverage of 8=0.7. Many layers of nickel metal may be attacked. The amount of surface accessible to hydrogen adsorption is considerably less than that expected on the basis of a simple ‘blocking mechanism’; this leads to the postulate of an arsenic atom being bonded to three nickel atoms while at the same time interacting with a second group of nickel sites sufficiently strongly to cause them to become inaccessible to chemisorption (but not enough to prevent them from participating in collective ferromagnetism). This finding seems to be the first piece of experimental evidence that a poison atom may exert its influence over host atoms differentially. An czAe value as high as 8, found after heat treatment at ca. ZOO”C, suggests the formation of clusters of AsNi, with n being as high as 12. Such clusters may change to As,Ni, when the amount of arsenic increases in the nickel catalyst. The electronic structure of the AsNi,, cluster is expected to be very similar to that of Ni,,, which was treated by Messmer et al. In the light of the present work, the cluster model approach might provide a unifying basis for the understanding high of the magnetic effect of carbon, sulphur and arsenic on nickel catalysts that are subject to heat treatment. Keywords: arsine poisoning, magnetic method, nickel/silica,
adsorption, cluster model.
INTRODUCTION
While sulphur poisoning has been the subject of extensive investigation in recent years, arsenic poisoning studies are relatively scarce, despite its relevance in industrial catalytic processes. In this paper we report our findings on the study of the arsine poisoning of Ni/SiO, catalysts and the subsequent chemisorption of hydrogen using the Weiss extraction method. The results are
0166-9834/91/$03.50
0 1991 Elsevier Science Publishers
B.V.
214
also discussed in the light of our earlier findings on the hydrogen sulphide poisoning of nickel catalysts. EXPERIMENTAL
The method of preparation of Ni/SiO, catalysts, the apparatus set-up and the procedure for adsorption and desorption experiments were similar to those outlined in a previous paper [ 11. The average nickel loading value and nickel particle diameter in the catalysts so obtained were lo-12% and 5.3 nm, respectively. Arsine was prepared by reducing As (OH), (aq) by potassium tetrahydroborate according to the method of Jolly and Drake [ 21. Magnetic data were obtained with an automatic Weiss extraction apparatus which has been described in detail elsewhere [ 31 s The value of bond number n was calculated by dividing ~1,the change in saturation magnetization per adsorbed molecule or atom, by the magnetic moment of nickel atoms, i.e. n= [B.M./molecule] / 0.606 B.M. RESULTS
AND DISCUSSION
Nature of chemical state of arsine The shape of the magnetic isotherm of arsine adsorption on Ni/SiOz shown in Fig. 1 suggests that the chemical nature of adsorbed arsine changes at different coverages. Obviously, three regions may be recognized. In region I, V,,, = O-8 ml, there is a marked decrease of magnetization of the catalyst. Although this part of the isotherm is somewhat curved, it is still possible to estimate an average value of m 4.0 B.M. - equivalent to a bond number of 6 by drawing the dotted line as shown. In region III, viz. V,,, = 17-36 ml, a very flat decrease in magnetization is effected by arsine adsorption. The region between the two could be considered to be a transition region. It is also interesting to point out that hydrogen evolution was observed beyond point E and that point M ( VA~H~ = 15 ml) corresponds to monolayer coverage when compared with the results of earlier hydrogen adsorption studies [ 1,5]. These results are therefore consistent with an initial dissociation adsorption of arsine with arsenic probably being bonded to three nickel sites, while the three hydrogen atoms are bonded to another three nickel sites, as shown in the following sketch: , N’l
/
As \ : NI
H \
\
\ Ni
H
H (1)
Nl
Ni
Ni
From a thermodynamic point of view, this model is well supported since the enthalpy change for the reaction ASH, + NiLNiAs + 3.2 H2 is - 278.7 kJ/mol. That desorption of arsine still takes place beyond point M indicates that ar-
215
\B c
A
A
30-
-
-_
t
I
0
I
4
1
8
1
12
1
16 20
1
24
1
28 32 36
I
I
1
40
Fig. 1. Change in magnetization as a function of adsorbed volume at room temperature for arsine. Curve A: for H,; Curve B is based on hydrogen adsorption results of Martin et al. [ 5 1,whose slope yields a”2 = 1.4 B.M./molecule. Point M represents monolayer coverage. Point E represents the volume adsorbed
at which hydrogen gas evolution
occurs.
senic has attacked more than the top layer of nickel atoms, although the bond number due to amine is apparently much lower when deeper layers are involved. This phenomenon is qualitatively very similar to that found by Ng and Martin [l] for hydrogen sulphide adsorption, where it was deduced that hydrogen sulphide was bonded to four nickel atoms of the first layer and approximately to two nickel atoms of the second layer. There are differences between arsine and hydrogen sulphide adsorption, however, in that the former has a transition region and hydrogen evolves much earlier - before monolayer coverage is reached ( tlAAsH3 = 0.7 ) , whereas in the latter case virtually no transition region is present and hydrogen evolution occurs only after two layers of nickel are covered (&,, = 2.0). Furthermore, the very existence of the transition region and the fact t.hat hydrogen evolution does not occur until half-way through this region imply that some less dissociated species, such as
216
H I
H\\ or
AS\
/, \
H
As'
\
Ni”
‘Ni (II)
Nl (III)
may be present when more than half a monolayer is covered by arsine. Structures II and III have been suggested earlier by Al-Daher and Saleh [ 41 in their studies of arsine adsorption on metallic nickel films at - 80’ C. However, it should be pointed out that their deduction was based on the assumption that only the very top layer of the nickel surface would be attacked. Finally, arsine adsorption differs from that of hydrogen sulphide in one more aspect: adsorption of arsine was found to occur even after the magnetization value of the nickel catalyst sample dropped to zero (not shown in Fig. 1) suggesting that much deeper layers were involved, leading to the formation of bulk Ni,As, which does not possess a superparamagnetic property. Nature of chemical states of arsenic in (on) nickel Curves B and C in Fig. 2 shows that the volume of hydrogen released at a given temperature is higher with greater arsine coverage. If we take curve A into consideration as well, it appears that hydrogen evolves more effectively on an arsine covered surface than on a hydrogen covered surface. Futhermore, all hydrogen could be completely removed from the surface at 673 K. These observations are quite different from that of the hydrogen sulphide case, where the presence of sulphur suppresses the release of hydrogen and a considerable amount of hydrogen is retained in (on) the poisoned catalyst, even at 773 K. Thus it appears that, in contrast to sulphur, arsenic has a promoting effect on hydrogen evolution. Alternatively speaking, arsenic weakens the Ni-H bond and this therefore implies that the effect of arsenic is not confined to the sites to which it is directly bonded. This deduction is further supported by the results of hydrogen adsorption on a poisoned surface (Fig. 5). That the an2 value remains almost the same as that for a bare nickel surface up to 8 ml [NTP]/g Ni of arsine coverage, as shown in Fig. 3, and the straight line yielding an2 = 1.3 for hydrogen readsorption, shown in Fig. 4, indicate that the uncovered part of a partially arsine poisoned surface accessible to hydrogen acts as a ‘free’ surface toward hydrogen adsorption. It would therefore be expected that, as in the case of hydrogen sulphide, the corresponding VH2vs* V.4, HB P lot should be a straight line up to the transition region (dotted line in Fig. 5 ) because only a blocking mechanism is involved. The experimentally observed curve falls short of this expectation. The maximum uptake is always lower than the value expected from a simple blocking mechanism and
217
Fig. 2. Amount of desorbed hydrogen
( Vn, per V nP adsorbed as Hz or as ASH,) from Ni/SiO,
with
various arsine coverages as a function of temperature [A: Ni covered with H,; B: Ni covered with 2.9 ml(NTP)AsH,/g Ni; C: Ni covered with 7.1 ml(NTP)AsH,/g Nil.
indeed almost reaches zero at 7-8 ml/g Ni amine coveragea and is completely zero at about 10 ml/g Ni ASH, coverage, a value that corresponds to the commencement of hydrogen evolution on an arsine covered surface. These experimental results can only be explained by postulating that on a partially arsine covered surface, some parts of the bare surface (though bare) are not accessible to further hydrogen adsorption and those parts which are accessible behave as ‘free’ surface as far as hydrogen adsorption is concerned. In fact, one can calculate the V,, vs. VA, H3 relationship on the basis of a model which envisages that each adsorbed arsenic atom, apart form directly binding three nickel sites, will interact fairly strongly with five more sites so that altogether eight sites will not be accessible to adsorption of a gas that requires free surface sites (although the five additional nickel atoms will still participate in collective ferromagnetism) ~ Such a theoretical relationship is shown as Curve C in Fig. 5 and should be compared with dashed line B which is expected on the basis of a simple blocking mechanism based on species I OThe value of i-8 ml NTP/g as carbon monoxide,
Ni is also the same value of coverage beyond which other gases such
ethylene. acetylene
or propene,
studied in this work, cannot be adsorbed.
218
t L
0
1
I
2
3
4
5
Fig. 3. cxH, vs. the volume of preadsorbed
6
7
8
ijA,E3(ml
NTP/gNi)
arsine.
throughout region I of arsine adsorption. (It should be noted that the shape of the dashed line B beyond the region I of arsine adsorption is ill defined.) The closeness between curves C and A appears to lend support to this model. Such a model is indeed in line with the thinking that a poison atom may have a detrimental influence beyond its nearest neighbours, which has been adopted by various authors in discussing gas adsorbate that requires multiple sites (see e.g. Andersen et al. [ 671) Martinet al. [ 8,9] ). To the knowledge of the present authors, our results are the first piece of direct experimental evidence which shows that, firstly a poison atom can exert an influence on two sets of atoms differentially and, secondly, the inaccessibility of the second set of atoms is definitely not due to any ‘steric effect’ since the adsorbate in question is hydrogen which does not require multiple sites. On the basis of this finding, one may expect that the poisoning effect of arsenic should be more severe than that of sulphur, at least for reactions involving hydrogen. This expectation is quite different from the results reported by Nielsen and Villadsen [lo]. The chemical states of arsenic in (on ) the nickel surface heat treatment is also of interest. Fig. 6 shows t,hat the cy value of a sample (after subtracting (xu due to remaining surface hydrogen, if any) is considerably higher after hydrogen evolution at elevated temperature (measurements were made after
219
Fig. 4. Change in magnetization
of a nickel catalyst treated with 6.6 ml of ax-sine at room temper-
ature and outgassed at 450’ C as a function an (Y value of 1.4 B.M./molecule.)
of
V,, readsorbed at room temperature.
(Slope yields
cooling down). An cyA,value as high as 8.0 (i.e. bond number = - 12-13) was observed for a sample of low arsine coverage and o!yAs was found to decrease as the amount of As increased. Such a high value is indeed astonishing because the arsenide with the highest Ni/As ratio reported is Ni5Asz. Furthermore, if one adopts the approach of Martin et al. [ 111 in interpreting the reduction of magnetic moment of a nickel catalyst by interstitial carbon, then the maximum number of electrons that can be donated to the d-band of metallic nickel by an interstitial arsine atom is five, i.e. the maximum ayAsvalue should be five, rather than the eight found experimentally. A more plausible explanation for the large observed bond number may be due to the formation of AsNils clusters, which results from the thermally induced migration of a surface arsenic atom during heat treatment to displace one of the bulk nickel atoms (most likely in the second or third layer below the surface). Imagine that this arsenic atom decreases the density of states at Fermi level energy (n(E,) ) of ca. 12 neighbouring nickel atoms in such a way that the well known Stoner criterion of ferromagnetism, UX n (I&.)> T (U,the exchange interaction energy; T the temperature in question) is not satisfied, then macroscopically, a decrease in magnetization will be observed which corresponds to 12 nickel atoms not par-
Fig. 5. Volume of adsorbed hydrogen at room temperature vs. the volume of preadsorbed amine. A: experimental data; B: expected curve if simple blocking mechanism operates with the same adsorbed species (Species I) throughout the first region as arsine adsorption. C: calculated on the basis of presently proposed model of an arsenic atom binding to three nickel sites and interacting fairly strongly with a second group of five nickel sites.
ticipating in collective ferromagnetism. Another approach to explaining the magnetic property arising from the AsNi,, cluster is based on the cluster model. Such a cluster is very similar to the cube-octahedral Nils cluster whose electronic structure has been calculated and fully discussed by Messmer et al. [ 121. According to these authors, amongst the spin polarized outer energy levels, 68 and 62 electrons of the nickel atoms (nickel has 10 outermost electrons, 3ds4s2) will occupy the spin-up and spin-down levels, respectively, leaving a total of six unpaired electrons and thereby leading to a net spin polarization and paramagnetism of the cluster. Now, in view of the close proximity of nickel and arsine in electronegativity and size (;cNi= 1.8, J&= 2.0; rr+Ji =0.125 nm, r&,=0.121 nm), it is conceivable that the substitution of the central nickel atom by an arsine atom is realistic and that the resulting AsNi,, cluster will presumably have an electronic structure very similar to that of a Ni,, cluster, except for the fact that the five extra electrons brought in by the arsenic atom ( 3d1°4s2p3) will occupy the spin-down levels, so that the AsNil cluster has only one paramagnetic spin, rather than six as in the case of Nil3 cluster. To a
221
90
80 . 70. 60. 50. 40. 30. 201 0
373
473
573
673
T( K)
Fig. 6. cy as a function of the desorption temperature for arsine. [A, B and C correspond, tively, to VAsH3= 1.3,3.8 and 8.2 ml NTP/g Ni.]
respec-
first approximation, the expected effect of the formation of such an AsNi,, cluster on magnetization of the nickel catalyst should be a reduction corresponding to 12 nickel atoms being hindered from participating in collective ferromagnetism. The experimental results are not far from this predicted reduction. It is interesting to note that, apart from the high magnetic effect reported for carbon, as mentioned above, Ng and Martin [1 ] also found that, at elevated temperature, a sulphur atom could decrease the magnetization of nickel by as much as 4 B.M. (corresponding to 6-7 nickel atoms being hindered from contributing to collective ferromagnetism). In the light of the present work, it appears that the cluster model might provide a unifying explanation for these phenomena. Both carbon and sulphur differ significantly from nickel in terms of electronegativity, size, and number of electrons in the outer shells and are therefore very unlikely to form an XNi,, (X = C, S ) cluster”. Instead, a XNis cluster may more likely be formed when the nickel catalyst (with X adsorbed) is thermally treated. Theoretical calculations of energy levels for AsNi,,, CNis and SNiGwould be very useful to confirm the feasibility of the proposed AsNil cluster and the general validity of the cluster model in interpreting the ob“For a close packed structure, it is possible to retain ligancy 12 with a central sphere as much as 10% smaller than the surrounding spheres. These spheres are then arranged at the corners of a regular icosahedron. See e.g. Pauling [ 131.
222
served high magnetic effects of carbon, sulphur and arsenic on nickel catalysts subjected to heat treatments. The heat treatment results obtained, as shown in Fig. 7, provide similar information. (Here the sample was not evacuated and the hydrogen desorbed should be readsorbed onto the surface after cooling.) Perhaps two more points should be noted. First, for a surface with greater amount of arsenic, both Fig. 6 and Fig. 7 show that the aAs value is much smaller, suggesting formation of cluster-like As,Ni, (or even bulk arsenides, as mentioned earlier) at high coverage. Second, in comparing Fig. 6 with Fig. 7, the presence of hydrogen has very little effect on the value of a& ;IZ agreement with our deduction in the discussion of hydrogen adsorption presented earlier. By comparison with our previous studies of hydrogen sulphide, one can see that the effect of As on catalytic reactions in general (such as partial hydrogenation of acetylene to ethylene, of polyalkenes or cyclic polyenes to corresponding monoalkenes) will be expected to be more severe than that of sulphur. In our next communication we shall report the effects of arsenic on the adsorption of carbon monoxide, acetylene, ethylene and propene. In comparing the results obtained for carbon monoxide and acetylene with those of hydrogen sulphide arsine effects are more complex, but no doubt render the sura AS~;3(B. M. /molecule)
373
473
c 573
Fig. 7.LYas a function of the holding temperature spectively, to VA, H3 =0.85, 1.8, 7.1 ml NTP/g Ni.
T( K)
673 for arsine. Curves A, B and C correspond
re-
223
face significantly less accessible to these gases (the ratio of maximum uptake of carbon monoxide for arsenic to sulphur is 1: 3 ). As regards acetylene, a completely sulphur poisoned nickel surface was found still to be capable of adsorbing, while a surface with - 7 ml AsH,/g Ni was found to be incapable of adsorption of acetylene and was therefore more ‘poisonous’ for reactions involving either the adsorbed carbon monoxide or acetylene species as precursor (or adsorption of carbon monoxide or acetylene as a rate determining step). Insofar as hydrocarbon cracking is concerned, in a broad sense, the inhibiting power of arsenic was found to be comparable with that of sulphur. However, the effect was again found more complex in the case of arsenic and this will be also discussed in our next communication [ 141. CONCLUSION
Arsenic can attack many layers of a nickel catalyst at room temperature. Before hydrogen evolution at 0% 0.7, the chemical state of arsine in (on) the nickel catalyst is probably H H H I AS\ \ I
N;
Ni
‘Xi
Ni
Ni
Ni
An arsenic atom, apart form bonding to a first group of nickel atoms, can interact with a second group of nickel sites causing them to be inaccessible to hydrogen adsorption. Such an interaction, however, is not strong enough to prevent the second group of nickel sites from participating in collective ferromagnetism. For those parts of the nickel surface which are available for hydrogen adsorption, they behave as free surface. This finding seems to be the first piece of experimental evidence that a poison atom may exert its influence over host atoms differentially. The chemical state of arsenic undergoes changes at elevated temperature. The very high aAs value observed when the amount of arsenic in a sample is small could be accounted for by postulating the formation of an AsNil cluster arising from thermally induced migration of a surface arsenic atom to replace a bulk nickel atom. The electronic structure of such a cluster is expected to be very similar to that of NilY cluster which has been studied by Messmer et al. [ 121. When the amount of arsenic increases, the number of nickel atoms participating in bonding decreases, suggesting formation of a cluster such as As,Ni, at higher coverage. The behaviour of the poisoned surface towards adsorption suggests that the poisoning effect of arsenic should be more severe than that of sulphur, at least for reactions involving hydrogen. ACKNOWLEDGEMENT
We should like to thank Messrs. C.P. Luk, F. Lam and Miss Jennifer Lie of the Department of Chemistry, University of Hong Kong for their assistance
224
in the technical work. One of us (Y.J. Chang) is indebted to the Pei Hua Education Foundation of Hong Kong for a Fellowship. Helpful comments from Dr. G.A. Martin are also gratefully acknowledged.
REFERENCES
2 3 4 5 6 7 8 9 LO 11 12 13 14
C.F. Ng and G.A. Martin, J. Catal., 54 (1978) 384. J.E. Drake and C. Riddle, J. Inorg. Synt., 13 (1957) 14. Y.J. Chang, C.F. Ng and W.K. Kwan, Rev. Sci. Instru., 59 (1988) 342. I.M. Al-Daher and J.M. SaIeh, J. Phys. Chem., 76 (19’72) 2851 and references therein. G.A. Martin, B. Imelik and M. Pettre, J. Chim. Phys., 66 (1969) 1682. N.T. Andersen, F. Topsee, I. Alstrup and J.R. Rostrup-Nielsen, J. Catal., 104 (1987) 454. I. Alstrup and N.T. Andersen, J. Catal., 104 (1987) 466. G.A. Martin, Catal. Rev. - Sci. Eng., 30 (1988) 519. G.A. Martin and CF. Ng, Appl. Catal., 31 (1987) 235. B. Nielsen and J. Villadsen, Appl. Cat& 11 (1984) 123. G.A. Martin, M. Primet and J.A. Dalmon, J. Catal., 53 (1978) 321. R.P. Messmer, S.K. Krmdson, K.J. Johnson, J.B. Diamond and C.Y. Yang, Phys. Rev. B, 13 (1976) 1396. L. Pauling, The Nature of the Chemical Bond, 3rd Ed., Cornell Univ. Press, 1980. C.F. Ng and Y.J. Chang, to be submitted to Appl. Catal.